Nvidia’s RTX 4060 Ti is the latest Lovelace GPU to be the subject of an info dump regarding its purported spec, as rumors about the RTX 4000 range continue to come thick and fast.
This speculation is from Kopite7kimi, a regular when it comes to Twitter hardware leaks, and it provides a fair few details (hat tip to VideoCardz (opens in new tab) for spotting the tweet).
RTX 4060 Ti has a very short reference board. The PG190 still uses CEM5 connector. AD106-350-A14352FP328G 18Gbps GDDR632M L2220WDecember 13, 2022
Go into this with plentiful seasoning in hand, but as you can see, the GPU used is AD106 (supposedly AD106-350-A1, to be precise), and we have a CUDA Core count of 4,352, alongside 8GB of GDDR6 video RAM (at 18Gbps). The TGP is pegged at 220W.
As VideoCardz points out, previous spillage has suggested that the AD106 GPU will run with a 128-bit memory bus, which would give the RTX 4060 Ti a bandwidth of 288GB/s in theory.
We’re also told that the RTX 4060 Ti has a “very short” reference board.
Analysis: Still a long road ahead for this GPU
The purported spec shown here has prompted disappointment as a general reaction online.
Naturally, we should be skeptical about any rumored specs – and we should bear in mind that this is still early development for the RTX 4060 Ti, as far as we’re aware, so even if true, the specifications floated here could still be changed by Nvidia down the line.
The CUDA Core count itself has drawn some groans, but we can’t judge the power of a card purely off that spec. Yes, it actually has fewer cores than the RTX 3060 Ti (which has 4,864) according to this spillage, but there are generational architectural improvements to take into account, and all the other GPU specs too.
What’s more of a sticking point for many is the use of 8GB of VRAM, and that rather meagre looking theoretical memory bandwidth, which could be an unfortunate combination – though the much-increased L2 cache, 32MB, could be a helpful boost compared to the 3060 Ti which had 4MB. Furthermore, we should add extra salt here, as the bandwidth calculation is based on a previous rumored spec, which makes things shakier.
At any rate, gamers out there were already unhappy with the 3060 Ti using 8GB, let alone a more powerful 4060 Ti. But whatever the case, broad expectations are for a graphics card that’s the equal of the RTX 3070 Ti – and a lot will then depend on the pricing. Which given Nvidia’s current form, doesn’t make us optimistic.
Above all, we should remember that all these specs could change based on the reaction to the RTX 4070 Ti and 4070, and how things pan out with mid-range RDNA 3 GPUs, of course. There’s likely a long time to go, too, before the RTX 4060 Ti emerges – certainly the rumor mill doesn’t believe the RTX 4060 will be out until mid-2023 – and Nvidia still has a fair whack of RTX 3000 stock to clear, as we all know, no doubt a good deal of it in the mid-range.
So, given that lengthy timeframe for the RTX 4060 Ti and 4060’s likely appearance, even if these are the specs on the board right now, we’d be surprised if they didn’t change at all over time based on the factors mentioned and changing GPU landscape.